arch/x86/include/asm/pgtable.h

   1 #ifndef _ASM_X86_PGTABLE_H
   2 #define _ASM_X86_PGTABLE_H
   3
   4 #include <asm/page.h>
   5 #include <asm/e820.h>
   6
   7 #include <asm/pgtable_types.h>
   8
   9 /*
  10  * Macro to mark a page protection value as UC-
  11  */
  12 #define pgprot_noncached(prot)                                  \
  13         ((boot_cpu_data.x86 > 3)                                \
  14          ? (__pgprot(pgprot_val(prot) | _PAGE_CACHE_UC_MINUS))  \
  15          : (prot))
  16
  17 #ifndef __ASSEMBLY__
  18
  19 #include <asm/x86_init.h>
  20
  21 /*
  22  * ZERO_PAGE is a global shared page that is always zero: used
  23  * for zero-mapped memory areas etc..
  24  */
  25 extern unsigned long empty_zero_page[PAGE_SIZE / sizeof(unsigned long)];
  26 #define ZERO_PAGE(vaddr) (virt_to_page(empty_zero_page))
  27
  28 extern spinlock_t pgd_lock;
  29 extern struct list_head pgd_list;
  30
  31 extern struct mm_struct *pgd_page_get_mm(struct page *page);
  32
  33 #ifdef CONFIG_PARAVIRT
  34 #include <asm/paravirt.h>
  35 #else  /* !CONFIG_PARAVIRT */
  36 #define set_pte(ptep, pte)              native_set_pte(ptep, pte)
  37 #define set_pte_at(mm, addr, ptep, pte) native_set_pte_at(mm, addr, ptep, pte)
  38 #define set_pmd_at(mm, addr, pmdp, pmd) native_set_pmd_at(mm, addr, pmdp, pmd)
  39
  40 #define set_pte_atomic(ptep, pte)                                       \
  41         native_set_pte_atomic(ptep, pte)
  42
  43 #define set_pmd(pmdp, pmd)              native_set_pmd(pmdp, pmd)
  44
  45 #ifndef __PAGETABLE_PUD_FOLDED
  46 #define set_pgd(pgdp, pgd)              native_set_pgd(pgdp, pgd)
  47 #define pgd_clear(pgd)                  native_pgd_clear(pgd)
  48 #endif
  49
  50 #ifndef set_pud
  51 # define set_pud(pudp, pud)             native_set_pud(pudp, pud)
  52 #endif
  53
  54 #ifndef __PAGETABLE_PMD_FOLDED
  55 #define pud_clear(pud)                  native_pud_clear(pud)
  56 #endif
  57
  58 #define pte_clear(mm, addr, ptep)       native_pte_clear(mm, addr, ptep)
  59 #define pmd_clear(pmd)                  native_pmd_clear(pmd)
  60
  61 #define pte_update(mm, addr, ptep)              do { } while (0)
  62 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
  63 #define pmd_update(mm, addr, ptep)              do { } while (0)
  64 #define pmd_update_defer(mm, addr, ptep)        do { } while (0)
  65
  66 #define pgd_val(x)      native_pgd_val(x)
  67 #define __pgd(x)        native_make_pgd(x)
  68
  69 #ifndef __PAGETABLE_PUD_FOLDED
  70 #define pud_val(x)      native_pud_val(x)
  71 #define __pud(x)        native_make_pud(x)
  72 #endif
  73
  74 #ifndef __PAGETABLE_PMD_FOLDED
  75 #define pmd_val(x)      native_pmd_val(x)
  76 #define __pmd(x)        native_make_pmd(x)
  77 #endif
  78
  79 #define pte_val(x)      native_pte_val(x)
  80 #define __pte(x)        native_make_pte(x)
  81
  82 #define arch_end_context_switch(prev)   do {} while(0)
  83
  84 #endif  /* CONFIG_PARAVIRT */
  85
  86 /*
  87  * The following only work if pte_present() is true.
  88  * Undefined behaviour if not..
  89  */
  90 static inline int pte_dirty(pte_t pte)
  91 {
  92         return pte_flags(pte) & _PAGE_DIRTY;
  93 }
  94
  95 static inline int pte_young(pte_t pte)
  96 {
  97         return pte_flags(pte) & _PAGE_ACCESSED;
  98 }
  99
 100 static inline int pte_write(pte_t pte)
 101 {
 102         return pte_flags(pte) & _PAGE_RW;
 103 }
 104
 105 static inline int pte_file(pte_t pte)
 106 {
 107         return pte_flags(pte) & _PAGE_FILE;
 108 }
 109
 110 static inline int pte_huge(pte_t pte)
 111 {
 112         return pte_flags(pte) & _PAGE_PSE;
 113 }
 114
 115 static inline int pte_global(pte_t pte)
 116 {
 117         return pte_flags(pte) & _PAGE_GLOBAL;
 118 }
 119
 120 static inline int pte_exec(pte_t pte)
 121 {
 122         return !(pte_flags(pte) & _PAGE_NX);
 123 }
 124
 125 static inline int pte_special(pte_t pte)
 126 {
 127         return pte_flags(pte) & _PAGE_SPECIAL;
 128 }
 129
 130 static inline unsigned long pte_pfn(pte_t pte)
 131 {
 132         return (pte_val(pte) & PTE_PFN_MASK) >> PAGE_SHIFT;
 133 }
 134
 135 static inline unsigned long pmd_pfn(pmd_t pmd)
 136 {
 137         return (pmd_val(pmd) & PTE_PFN_MASK) >> PAGE_SHIFT;
 138 }
 139
 140 #define pte_page(pte)   pfn_to_page(pte_pfn(pte))
 141
 142 static inline int pmd_large(pmd_t pte)
 143 {
 144         return (pmd_flags(pte) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 145                 (_PAGE_PSE | _PAGE_PRESENT);
 146 }
 147
 148 static inline pte_t pte_set_flags(pte_t pte, pteval_t set)
 149 {
 150         pteval_t v = native_pte_val(pte);
 151
 152         return native_make_pte(v | set);
 153 }
 154
 155 static inline pte_t pte_clear_flags(pte_t pte, pteval_t clear)
 156 {
 157         pteval_t v = native_pte_val(pte);
 158
 159         return native_make_pte(v & ~clear);
 160 }
 161
 162 static inline pte_t pte_mkclean(pte_t pte)
 163 {
 164         return pte_clear_flags(pte, _PAGE_DIRTY);
 165 }
 166
 167 static inline pte_t pte_mkold(pte_t pte)
 168 {
 169         return pte_clear_flags(pte, _PAGE_ACCESSED);
 170 }
 171
 172 static inline pte_t pte_wrprotect(pte_t pte)
 173 {
 174         return pte_clear_flags(pte, _PAGE_RW);
 175 }
 176
 177 static inline pte_t pte_mkexec(pte_t pte)
 178 {
 179         return pte_clear_flags(pte, _PAGE_NX);
 180 }
 181
 182 static inline pte_t pte_mkdirty(pte_t pte)
 183 {
 184         return pte_set_flags(pte, _PAGE_DIRTY);
 185 }
 186
 187 static inline pte_t pte_mkyoung(pte_t pte)
 188 {
 189         return pte_set_flags(pte, _PAGE_ACCESSED);
 190 }
 191
 192 static inline pte_t pte_mkwrite(pte_t pte)
 193 {
 194         return pte_set_flags(pte, _PAGE_RW);
 195 }
 196
 197 static inline pte_t pte_mkhuge(pte_t pte)
 198 {
 199         return pte_set_flags(pte, _PAGE_PSE);
 200 }
 201
 202 static inline pte_t pte_clrhuge(pte_t pte)
 203 {
 204         return pte_clear_flags(pte, _PAGE_PSE);
 205 }
 206
 207 static inline pte_t pte_mkglobal(pte_t pte)
 208 {
 209         return pte_set_flags(pte, _PAGE_GLOBAL);
 210 }
 211
 212 static inline pte_t pte_clrglobal(pte_t pte)
 213 {
 214         return pte_clear_flags(pte, _PAGE_GLOBAL);
 215 }
 216
 217 static inline pte_t pte_mkspecial(pte_t pte)
 218 {
 219         return pte_set_flags(pte, _PAGE_SPECIAL);
 220 }
 221
 222 /*
 223  * Mask out unsupported bits in a present pgprot.  Non-present pgprots
 224  * can use those bits for other purposes, so leave them be.
 225  */
 226 static inline pgprotval_t massage_pgprot(pgprot_t pgprot)
 227 {
 228         pgprotval_t protval = pgprot_val(pgprot);
 229
 230         if (protval & _PAGE_PRESENT)
 231                 protval &= __supported_pte_mask;
 232
 233         return protval;
 234 }
 235
 236 static inline pte_t pfn_pte(unsigned long page_nr, pgprot_t pgprot)
 237 {
 238         return __pte(((phys_addr_t)page_nr << PAGE_SHIFT) |
 239                      massage_pgprot(pgprot));
 240 }
 241
 242 static inline pmd_t pfn_pmd(unsigned long page_nr, pgprot_t pgprot)
 243 {
 244         return __pmd(((phys_addr_t)page_nr << PAGE_SHIFT) |
 245                      massage_pgprot(pgprot));
 246 }
 247
 248 static inline pte_t pte_modify(pte_t pte, pgprot_t newprot)
 249 {
 250         pteval_t val = pte_val(pte);
 251
 252         /*
 253          * Chop off the NX bit (if present), and add the NX portion of
 254          * the newprot (if present):
 255          */
 256         val &= _PAGE_CHG_MASK;
 257         val |= massage_pgprot(newprot) & ~_PAGE_CHG_MASK;
 258
 259         return __pte(val);
 260 }
 261
 262 /* mprotect needs to preserve PAT bits when updating vm_page_prot */
 263 #define pgprot_modify pgprot_modify
 264 static inline pgprot_t pgprot_modify(pgprot_t oldprot, pgprot_t newprot)
 265 {
 266         pgprotval_t preservebits = pgprot_val(oldprot) & _PAGE_CHG_MASK;
 267         pgprotval_t addbits = pgprot_val(newprot);
 268         return __pgprot(preservebits | addbits);
 269 }
 270
 271 #define pte_pgprot(x) __pgprot(pte_flags(x) & PTE_FLAGS_MASK)
 272
 273 #define canon_pgprot(p) __pgprot(massage_pgprot(p))
 274
 275 static inline int is_new_memtype_allowed(u64 paddr, unsigned long size,
 276                                          unsigned long flags,
 277                                          unsigned long new_flags)
 278 {
 279         /*
 280          * PAT type is always WB for untracked ranges, so no need to check.
 281          */
 282         if (x86_platform.is_untracked_pat_range(paddr, paddr + size))
 283                 return 1;
 284
 285         /*
 286          * Certain new memtypes are not allowed with certain
 287          * requested memtype:
 288          * - request is uncached, return cannot be write-back
 289          * - request is write-combine, return cannot be write-back
 290          */
 291         if ((flags == _PAGE_CACHE_UC_MINUS &&
 292              new_flags == _PAGE_CACHE_WB) ||
 293             (flags == _PAGE_CACHE_WC &&
 294              new_flags == _PAGE_CACHE_WB)) {
 295                 return 0;
 296         }
 297
 298         return 1;
 299 }
 300
 301 pmd_t *populate_extra_pmd(unsigned long vaddr);
 302 pte_t *populate_extra_pte(unsigned long vaddr);
 303 #endif  /* __ASSEMBLY__ */
 304
 305 #ifdef CONFIG_X86_32
 306 # include "pgtable_32.h"
 307 #else
 308 # include "pgtable_64.h"
 309 #endif
 310
 311 #ifndef __ASSEMBLY__
 312 #include <linux/mm_types.h>
 313
 314 static inline int pte_none(pte_t pte)
 315 {
 316         return !pte.pte;
 317 }
 318
 319 #define __HAVE_ARCH_PTE_SAME
 320 static inline int pte_same(pte_t a, pte_t b)
 321 {
 322         return a.pte == b.pte;
 323 }
 324
 325 static inline int pte_present(pte_t a)
 326 {
 327         return pte_flags(a) & (_PAGE_PRESENT | _PAGE_PROTNONE);
 328 }
 329
 330 static inline int pte_hidden(pte_t pte)
 331 {
 332         return pte_flags(pte) & _PAGE_HIDDEN;
 333 }
 334
 335 static inline int pmd_present(pmd_t pmd)
 336 {
 337         return pmd_flags(pmd) & _PAGE_PRESENT;
 338 }
 339
 340 static inline int pmd_none(pmd_t pmd)
 341 {
 342         /* Only check low word on 32-bit platforms, since it might be
 343            out of sync with upper half. */
 344         return (unsigned long)native_pmd_val(pmd) == 0;
 345 }
 346
 347 static inline unsigned long pmd_page_vaddr(pmd_t pmd)
 348 {
 349         return (unsigned long)__va(pmd_val(pmd) & PTE_PFN_MASK);
 350 }
 351
 352 /*
 353  * Currently stuck as a macro due to indirect forward reference to
 354  * linux/mmzone.h's __section_mem_map_addr() definition:
 355  */
 356 #define pmd_page(pmd)   pfn_to_page(pmd_val(pmd) >> PAGE_SHIFT)
 357
 358 /*
 359  * the pmd page can be thought of an array like this: pmd_t[PTRS_PER_PMD]
 360  *
 361  * this macro returns the index of the entry in the pmd page which would
 362  * control the given virtual address
 363  */
 364 static inline unsigned long pmd_index(unsigned long address)
 365 {
 366         return (address >> PMD_SHIFT) & (PTRS_PER_PMD - 1);
 367 }
 368
 369 /*
 370  * Conversion functions: convert a page and protection to a page entry,
 371  * and a page entry and page directory to the page they refer to.
 372  *
 373  * (Currently stuck as a macro because of indirect forward reference
 374  * to linux/mm.h:page_to_nid())
 375  */
 376 #define mk_pte(page, pgprot)   pfn_pte(page_to_pfn(page), (pgprot))
 377
 378 /*
 379  * the pte page can be thought of an array like this: pte_t[PTRS_PER_PTE]
 380  *
 381  * this function returns the index of the entry in the pte page which would
 382  * control the given virtual address
 383  */
 384 static inline unsigned long pte_index(unsigned long address)
 385 {
 386         return (address >> PAGE_SHIFT) & (PTRS_PER_PTE - 1);
 387 }
 388
 389 static inline pte_t *pte_offset_kernel(pmd_t *pmd, unsigned long address)
 390 {
 391         return (pte_t *)pmd_page_vaddr(*pmd) + pte_index(address);
 392 }
 393
 394 static inline int pmd_bad(pmd_t pmd)
 395 {
 396         return (pmd_flags(pmd) & ~_PAGE_USER) != _KERNPG_TABLE;
 397 }
 398
 399 static inline unsigned long pages_to_mb(unsigned long npg)
 400 {
 401         return npg >> (20 - PAGE_SHIFT);
 402 }
 403
 404 #define io_remap_pfn_range(vma, vaddr, pfn, size, prot) \
 405         remap_pfn_range(vma, vaddr, pfn, size, prot)
 406
 407 #if PAGETABLE_LEVELS > 2
 408 static inline int pud_none(pud_t pud)
 409 {
 410         return native_pud_val(pud) == 0;
 411 }
 412
 413 static inline int pud_present(pud_t pud)
 414 {
 415         return pud_flags(pud) & _PAGE_PRESENT;
 416 }
 417
 418 static inline unsigned long pud_page_vaddr(pud_t pud)
 419 {
 420         return (unsigned long)__va((unsigned long)pud_val(pud) & PTE_PFN_MASK);
 421 }
 422
 423 /*
 424  * Currently stuck as a macro due to indirect forward reference to
 425  * linux/mmzone.h's __section_mem_map_addr() definition:
 426  */
 427 #define pud_page(pud)           pfn_to_page(pud_val(pud) >> PAGE_SHIFT)
 428
 429 /* Find an entry in the second-level page table.. */
 430 static inline pmd_t *pmd_offset(pud_t *pud, unsigned long address)
 431 {
 432         return (pmd_t *)pud_page_vaddr(*pud) + pmd_index(address);
 433 }
 434
 435 static inline int pud_large(pud_t pud)
 436 {
 437         return (pud_val(pud) & (_PAGE_PSE | _PAGE_PRESENT)) ==
 438                 (_PAGE_PSE | _PAGE_PRESENT);
 439 }
 440
 441 static inline int pud_bad(pud_t pud)
 442 {
 443         return (pud_flags(pud) & ~(_KERNPG_TABLE | _PAGE_USER)) != 0;
 444 }
 445 #else
 446 static inline int pud_large(pud_t pud)
 447 {
 448         return 0;
 449 }
 450 #endif  /* PAGETABLE_LEVELS > 2 */
 451
 452 #if PAGETABLE_LEVELS > 3
 453 static inline int pgd_present(pgd_t pgd)
 454 {
 455         return pgd_flags(pgd) & _PAGE_PRESENT;
 456 }
 457
 458 static inline unsigned long pgd_page_vaddr(pgd_t pgd)
 459 {
 460         return (unsigned long)__va((unsigned long)pgd_val(pgd) & PTE_PFN_MASK);
 461 }
 462
 463 /*
 464  * Currently stuck as a macro due to indirect forward reference to
 465  * linux/mmzone.h's __section_mem_map_addr() definition:
 466  */
 467 #define pgd_page(pgd)           pfn_to_page(pgd_val(pgd) >> PAGE_SHIFT)
 468
 469 /* to find an entry in a page-table-directory. */
 470 static inline unsigned long pud_index(unsigned long address)
 471 {
 472         return (address >> PUD_SHIFT) & (PTRS_PER_PUD - 1);
 473 }
 474
 475 static inline pud_t *pud_offset(pgd_t *pgd, unsigned long address)
 476 {
 477         return (pud_t *)pgd_page_vaddr(*pgd) + pud_index(address);
 478 }
 479
 480 static inline int pgd_bad(pgd_t pgd)
 481 {
 482         return (pgd_flags(pgd) & ~_PAGE_USER) != _KERNPG_TABLE;
 483 }
 484
 485 static inline int pgd_none(pgd_t pgd)
 486 {
 487         return !native_pgd_val(pgd);
 488 }
 489 #endif  /* PAGETABLE_LEVELS > 3 */
 490
 491 #endif  /* __ASSEMBLY__ */
 492
 493 /*
 494  * the pgd page can be thought of an array like this: pgd_t[PTRS_PER_PGD]
 495  *
 496  * this macro returns the index of the entry in the pgd page which would
 497  * control the given virtual address
 498  */
 499 #define pgd_index(address) (((address) >> PGDIR_SHIFT) & (PTRS_PER_PGD - 1))
 500
 501 /*
 502  * pgd_offset() returns a (pgd_t *)
 503  * pgd_index() is used get the offset into the pgd page's array of pgd_t's;
 504  */
 505 #define pgd_offset(mm, address) ((mm)->pgd + pgd_index((address)))
 506 /*
 507  * a shortcut which implies the use of the kernel's pgd, instead
 508  * of a process's
 509  */
 510 #define pgd_offset_k(address) pgd_offset(&init_mm, (address))
 511
 512
 513 #define KERNEL_PGD_BOUNDARY     pgd_index(PAGE_OFFSET)
 514 #define KERNEL_PGD_PTRS         (PTRS_PER_PGD - KERNEL_PGD_BOUNDARY)
 515
 516 #ifndef __ASSEMBLY__
 517
 518 extern int direct_gbpages;
 519
 520 /* local pte updates need not use xchg for locking */
 521 static inline pte_t native_local_ptep_get_and_clear(pte_t *ptep)
 522 {
 523         pte_t res = *ptep;
 524
 525         /* Pure native function needs no input for mm, addr */
 526         native_pte_clear(NULL, 0, ptep);
 527         return res;
 528 }
 529
 530 static inline void native_set_pte_at(struct mm_struct *mm, unsigned long addr,
 531                                      pte_t *ptep , pte_t pte)
 532 {
 533         native_set_pte(ptep, pte);
 534 }
 535
 536 static inline void native_set_pmd_at(struct mm_struct *mm, unsigned long addr,
 537                                      pmd_t *pmdp , pmd_t pmd)
 538 {
 539         native_set_pmd(pmdp, pmd);
 540 }
 541
 542 #ifndef CONFIG_PARAVIRT
 543 /*
 544  * Rules for using pte_update - it must be called after any PTE update which
 545  * has not been done using the set_pte / clear_pte interfaces.  It is used by
 546  * shadow mode hypervisors to resynchronize the shadow page tables.  Kernel PTE
 547  * updates should either be sets, clears, or set_pte_atomic for P->P
 548  * transitions, which means this hook should only be called for user PTEs.
 549  * This hook implies a P->P protection or access change has taken place, which
 550  * requires a subsequent TLB flush.  The notification can optionally be delayed
 551  * until the TLB flush event by using the pte_update_defer form of the
 552  * interface, but care must be taken to assure that the flush happens while
 553  * still holding the same page table lock so that the shadow and primary pages
 554  * do not become out of sync on SMP.
 555  */
 556 #define pte_update(mm, addr, ptep)              do { } while (0)
 557 #define pte_update_defer(mm, addr, ptep)        do { } while (0)
 558 #endif
 559
 560 /*
 561  * We only update the dirty/accessed state if we set
 562  * the dirty bit by hand in the kernel, since the hardware
 563  * will do the accessed bit for us, and we don't want to
 564  * race with other CPU's that might be updating the dirty
 565  * bit at the same time.
 566  */
 567 struct vm_area_struct;
 568
 569 #define  __HAVE_ARCH_PTEP_SET_ACCESS_FLAGS
 570 extern int ptep_set_access_flags(struct vm_area_struct *vma,
 571                                  unsigned long address, pte_t *ptep,
 572                                  pte_t entry, int dirty);
 573
 574 #define __HAVE_ARCH_PTEP_TEST_AND_CLEAR_YOUNG
 575 extern int ptep_test_and_clear_young(struct vm_area_struct *vma,
 576                                      unsigned long addr, pte_t *ptep);
 577
 578 #define __HAVE_ARCH_PTEP_CLEAR_YOUNG_FLUSH
 579 extern int ptep_clear_flush_young(struct vm_area_struct *vma,
 580                                   unsigned long address, pte_t *ptep);
 581
 582 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR
 583 static inline pte_t ptep_get_and_clear(struct mm_struct *mm, unsigned long addr,
 584                                        pte_t *ptep)
 585 {
 586         pte_t pte = native_ptep_get_and_clear(ptep);
 587         pte_update(mm, addr, ptep);
 588         return pte;
 589 }
 590
 591 #define __HAVE_ARCH_PTEP_GET_AND_CLEAR_FULL
 592 static inline pte_t ptep_get_and_clear_full(struct mm_struct *mm,
 593                                             unsigned long addr, pte_t *ptep,
 594                                             int full)
 595 {
 596         pte_t pte;
 597         if (full) {
 598                 /*
 599                  * Full address destruction in progress; paravirt does not
 600                  * care about updates and native needs no locking
 601                  */
 602                 pte = native_local_ptep_get_and_clear(ptep);
 603         } else {
 604                 pte = ptep_get_and_clear(mm, addr, ptep);
 605         }
 606         return pte;
 607 }
 608
 609 #define __HAVE_ARCH_PTEP_SET_WRPROTECT
 610 static inline void ptep_set_wrprotect(struct mm_struct *mm,
 611                                       unsigned long addr, pte_t *ptep)
 612 {
 613         clear_bit(_PAGE_BIT_RW, (unsigned long *)&ptep->pte);
 614         pte_update(mm, addr, ptep);
 615 }
 616
 617 #define flush_tlb_fix_spurious_fault(vma, address)
 618
 619 /*
 620  * clone_pgd_range(pgd_t *dst, pgd_t *src, int count);
 621  *
 622  *  dst - pointer to pgd range anwhere on a pgd page
 623  *  src - ""
 624  *  count - the number of pgds to copy.
 625  *
 626  * dst and src can be on the same page, but the range must not overlap,
 627  * and must not cross a page boundary.
 628  */
 629 static inline void clone_pgd_range(pgd_t *dst, pgd_t *src, int count)
 630 {
 631        memcpy(dst, src, count * sizeof(pgd_t));
 632 }
 633
 634
 635 #include <asm-generic/pgtable.h>
 636 #endif  /* __ASSEMBLY__ */
 637
 638 #endif /* _ASM_X86_PGTABLE_H */